Here's a dirty example of a quick cloud lighting comparison test I have just run. All examples are on a default layer, 2000m thick.
GI is on in the atmosphere and there are no fill lights, for all tests.
Firstly, no GI in the clouds;
[attachimg=#]
Secondly, GI in clouds(default settings);
[attachimg=#]
Finally, no GI in the clouds, edited layer.
[attachimg=#]
It isn't a perfect representation, you can notice the difference all over with these back to back comparisons but, it is a nice enough 'fake' GI in clouds without using fill lights that could easily blow-out many other scene features. It may be a little too much here as the horizon is a bit over-bright in the edited layer, I think, but, meh! Close enough for government work, as they say.
The effect is more useful in dark scenes, where you really have more control, this test just tries to approximate it in a daytime setting to satisfy my own curiosity, really.
Night is much better because the measures need to be less subtle and they can be ridiculously extreme(as they are in the noctilucents of this thread) than they are here, you can make a really nice cloud lighting without GI in a night scene.
It isn't a ground-breaking method, it's completely simple and obvious, I just never thought to do it before. More later, in the dark...